國立虎尾科技大學 |

High Performance Computing and Real Time Software for High Dimensional Data Classification.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	High Performance Computing and Real Time Software for High Dimensional Data Classification./
Author:	Meng, Zhaoyi.
Description:	1 online resource (108 pages)
Notes:	Source: Dissertation Abstracts International, Volume: 79-07(E), Section: B.
Contained By:	Dissertation Abstracts International79-07B(E).
Subject:	Applied mathematics. -
Online resource:	click for full text (PQDT)
ISBN:	9780355623666

High Performance Computing and Real Time Software for High Dimensional Data Classification.
Meng, Zhaoyi.

High Performance Computing and Real Time Software for High Dimensional Data Classification. - 1 online resource (108 pages)

Source: Dissertation Abstracts International, Volume: 79-07(E), Section: B.

Thesis (Ph.D.)--University of California, Los Angeles, 2018.

Includes bibliographical references

We present high performance computing and real time software for high dimensional data classification. We investigate the OpenMP parallelization and optimization of two graph-based data classification algorithms. The new algorithms are based on graph and PDE solution techniques and provide significant accuracy and performance advantages over traditional data classification algorithm. We use OpenMP as the parallelization language to parallelize the most time-consuming parts, which is the Nystr\"om extension eigensolver. The Nystr\"om extension calculates eigenvalue/eigenvectors of the graph Laplacian and this is a self-contained module that can be used in conjunction with other graph-Laplacian based methods such as spectral clustering. We then optimize the OpenMP implementations and detail the performance on traditional supercomputer nodes (in our case a Cray XC30), and test the optimization steps on emerging testbed systems based on Intel's Knights Corner and Landing processors. We show both performance improvement and strong scaling behavior. A large number of optimization techniques and analyses are necessary before the algorithm reaches almost ideal scaling. We further develop the parallelized real time software. The software is published on Image Processing On Line (IPOL) and uses IPOL's server. It provides real time computation for image classification.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2018

Mode of access: World Wide Web

ISBN: 9780355623666Subjects--Topical Terms:

1069907
Applied mathematics.
Index Terms--Genre/Form:

554714
Electronic books.

High Performance Computing and Real Time Software for High Dimensional Data Classification.
LDR:04403ntm a2200361Ki 4500 001 916625
005 20180927111920.5
006 m o u
007 cr mn||||a|a||
008 190606s2018 xx obm 000 0 eng d
020 $a 9780355623666
035 $a (MiAaPQ)AAI10744314
035 $a (MiAaPQ)ucla:16536
035 $a AAI10744314
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Meng, Zhaoyi. $3 1190421
245 1 0 $a High Performance Computing and Real Time Software for High Dimensional Data Classification.
264 0 $c 2018
300 $a 1 online resource (108 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertation Abstracts International, Volume: 79-07(E), Section: B.
500 $a Adviser: Andrea L. Bertozzi.
502 $a Thesis (Ph.D.)--University of California, Los Angeles, 2018.
504 $a Includes bibliographical references
520 $a We present high performance computing and real time software for high dimensional data classification. We investigate the OpenMP parallelization and optimization of two graph-based data classification algorithms. The new algorithms are based on graph and PDE solution techniques and provide significant accuracy and performance advantages over traditional data classification algorithm. We use OpenMP as the parallelization language to parallelize the most time-consuming parts, which is the Nystr\"om extension eigensolver. The Nystr\"om extension calculates eigenvalue/eigenvectors of the graph Laplacian and this is a self-contained module that can be used in conjunction with other graph-Laplacian based methods such as spectral clustering. We then optimize the OpenMP implementations and detail the performance on traditional supercomputer nodes (in our case a Cray XC30), and test the optimization steps on emerging testbed systems based on Intel's Knights Corner and Landing processors. We show both performance improvement and strong scaling behavior. A large number of optimization techniques and analyses are necessary before the algorithm reaches almost ideal scaling. We further develop the parallelized real time software. The software is published on Image Processing On Line (IPOL) and uses IPOL's server. It provides real time computation for image classification.
520 $a We apply the parallelized classification algorithms to various problems. The first one is hyperspectral image classification, a challenging modality due to the dimension of pixels which can range from hundreds to over a thousand frequencies depending on the sensor. Our methods use the full dimensionality of the data and consider a similarity graph based on pairwise comparisons of pixels. The graph is segmented using a pseudospectral algorithm for graph clustering that requires information about the eigenfunctions of the graph Laplacian but does not require the computation of the full graph. The parallelized Nystr\"om extension method randomly samples the graph to construct a low rank approximation of the graph Laplacian. With at most a few hundreds eigenfunctions, we can implement the clustering method to solve variational problems for graph-cut-based semi-supervised and unsupervised classification problems. The second application is ego-motion classification of body-worn videos. Portable cameras record dynamic first-person video footage and these videos contain information on the motion of the individual on whom the camera is mounted, defined as ego. We address the task of discovering ego-motion from the video itself, without other external calibration information. We investigate the use of similarity transformations between successive video frames to extract signals reflecting ego-motions and their frequencies. We use the parallelized graph-based unsupervised and semi-supervised learning algorithms to segment the video frames into different ego-motion categories. We show very accurate results on both choreographed test videos and ego-motion videos provided by the Los Angeles Police Department.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2018
538 $a Mode of access: World Wide Web
650 4 $a Applied mathematics. $3 1069907
650 4 $a Computer science. $3 573171
650 4 $a Information science. $3 561178
655 7 $a Electronic books. $2 local $3 554714
690 $a 0364
690 $a 0984
690 $a 0723
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a University of California, Los Angeles. $b Mathematics. $3 1180661
773 0 $t Dissertation Abstracts International $g 79-07B(E).
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10744314 $z click for full text (PQDT)